China-Europe rail: Erenhot central corridor tops 3,500 trips, 3.32m tonnes, +25% YoY

The Analysis

Erenhot has a wind that gets into everything, and on 21 September it was doing what it always does while the 1,293rd China-Europe train pulled out and pushed the central corridor past 3,500 trips for the year. By 20 September the port had handled 3,320,800 tonnes. Volume up 25.11 per cent on the year. Trips up 8.09 per cent.

The corridor runs 76 routes into more than 70 hubs across 10-plus countries and has settled into a rhythm of 13 trains a day. New energy vehicles, solar modules and electronics are over 40 per cent of the volume. That is the version of the story everybody has already read, and all of it is true.

The version I would act on is the gap between 25.11 and 8.09. This corridor did not grow this year because somebody laid more track. It grew because each train carried more. If you are booking space on it, that difference decides where your container sits in the queue in December, and it is not the number in the headline.

I have run boxes over this line for years, and the first thing I tell anyone who asks about Erenhot is that it is not a station, it is a seam. The town sits on the border with Mongolia. The train leaves the Chinese yard, runs a short distance, and arrives at Zamyn-Uud on the other side, where the gauge changes. Chinese track and Mongolian track are different widths, so either the bogies get swapped under the wagon or the boxes come off and go onto a different platform. Everything good you have heard about this corridor is true on the running sections. At the seam, it is negotiable.

So read the two growth numbers as one number. Divide the tonnage by the trips: 3,320,800 tonnes over 3,500 trips is roughly 949 tonnes a train. Then divide the volume index by the trip index, 1.2511 over 1.0809, and you get about 1.16. The average train is carrying close to 16 per cent more than a year ago, while the number of trains grew 8.09 per cent. The growth this year is load, not frequency.

Load is the cheap way to grow, right up to the point where it is not. You can keep adding to a train until you meet the axle limit, the container weight limit, or the yard's ability to lift. A heavier average does not take a longer crane slot, but it does mean less tolerance for the badly loaded wagon, and it moves the queue to whoever is over on weight. Growth by load runs out quietly. Nobody announces the morning the average hits the ceiling; the yard simply starts weighing more carefully.

The rhythm arithmetic is worth doing, because everyone on this line has already done it. Thirteen trains a day across the 264 days to 21 September is about 3,432 departures, and the corridor actually did 3,500, so the year-to-date run rate has been slightly above 13. There are 101 days left in the year. At 13 a day that is another 1,313, which puts the year near 4,800 trips. That is arithmetic on the published numbers and not a forecast, but the forwarders on this corridor treat it as one, which is why space for the second half of November is being discussed in September.

Where the queue forms is a route question, and the answer is not the departure yard. The departure yard has been the easy part for a while. The queue forms at the seam, where the gauge changes and the crane hours are finite. Then along the Mongolian single-track sections, where nothing overtakes anything and one slow wagon sets the pace for the whole consist. Then at the Russian border, where an inspection can add days without adding information. Then at the second gauge change on the Polish side, at Malaszewicze. A corridor is only as fast as its slowest yard, and this one has two seams and a long thin middle.

The two seams fail in different ways, and that difference should change what you ask about. Erenhot and Zamyn-Uud is a border and a gauge change on a long, thin, single-track approach, so it fails on weather, on block sequencing and on how many trains arrived the same morning. The Polish seam sits inside a dense network where several corridors compete for the same transfer capacity, so it fails on volume rather than on weather. One is a road problem, the other is a junction problem, and a shipper who treats them as the same risk will buy the wrong buffer in the wrong place.

From November the yards slow down in a way that is hard to argue with. Fewer effective shifts, equipment that does not want to work in the cold, boxes handled twice because the first lift did not seat. The timetable is a promise about departures. It has never been a promise about arrivals, and the gap between the two widens by days rather than hours once the temperature drops. If your plan reads the departure schedule as the arrival schedule, you have built a plan for June.

For the exporter of new energy vehicles, solar modules or electronics, the corridor has earned its 40 per cent honestly. Those are goods where a day of transit has a number attached to it, where the value density justifies the premium, and where the packaging survives a gauge change and a crane. This is also why the tonnage outruns the trip count: a container of modules or battery packs is heavy, so a rising share of that cargo pushes the average weight per train up on its own. The mix change and the load change are the same event, seen from two ends.

For the shipper of dense, low-value cargo, my answer is no, and I would rather say it plainly than sell a lane. Rail is not for you, and putting a box of low-value goods on this corridor to diversify is paying for a risk you were not carrying. The corridor's own split tells you this. If more than 40 per cent is high-value, then most of it is not, and a good part of that majority is on the train for reasons that have nothing to do with the economics of the cargo: a contract commitment, a customer who wants the boxes early, a relationship worth more than the freight difference. Know which of those reasons is yours, because the reason decides whether you can leave when the arithmetic turns.

One more consequence of that mix, and it lands on the least valuable cargo first. When more than 40 per cent of the corridor's volume is high-value, slot allocation quietly follows the value, because that is where the service commitments and the claims exposure sit. If your cargo is in the other 60 per cent, you are not competing for space against the average shipper. You are competing against the average and losing to the top 40 per cent, which is why your confirmation arrives later every December and why nobody at the other end can explain it to you.

A small shipper does not buy trains, he buys a slot from a consolidator, and his real exposure is not the corridor average. It is the week when the big shippers take the capacity. In a quarter when the run rate is climbing, the consolidator serves the customer who committed volume first and the spot customer second, and that ordering is not personal. At thirty containers a month your leverage is being predictable, not being large. Book the same window every week, and after a while the operator starts treating your slot as a slot instead of as a favour.

The importer at the other end has a different problem, and it is the box rather than the train. Westbound and eastbound do not balance, and equipment does not teleport. When the westbound peak takes every available forty-foot unit, the eastbound shipper who needs boxes sits and waits for a repositioning that nobody promised him. That is a days-to-weeks problem, and it eats the transit advantage that made the lane attractive in the first place. If your flow runs eastbound, your question is not the departure schedule. It is where the empty units are on the day you need them.

The 76 routes and the 70-plus hubs are the part I find genuinely new. Ten years ago this was a corridor with a main line and a few sidings. Now it is a network with branches, and a network can go around a problem. If one gateway is congested, some volume moves to another. That optionality is real and thinner than it looks: the branches are not sized like the trunk, so you get one alternative and not three, and the alternative runs twice a week.

Which brings the first clock, and it is measured in weeks rather than days. Space for the second half of November is being discussed now. If your production finishes in the first week of December and the goods have to be in Europe before the retail freeze, that conversation is already late, and no amount of escalation creates a slot that has been sold. The rule I use is simple: for anything that must arrive before the customer's year-end, the train is booked before the goods are finished, and the goods are finished to the booking. The other order of events is how people end up air-freighting a third of a container in December.

A 21-day door-to-door plan is not 21 days of running, and that is where most plans go wrong. It is running days, plus dwell at the first seam, plus dwell at the second, plus whatever buffer you wrote down. Four boxes on a piece of paper; the running days are the one number you do not control, the two dwells are the ones you can measure, and the buffer is the only one you actually decide. Fill in the boxes from your last three shipments instead of from a quote, and the plan will hold in November. Fill them in from a quote, and you have written down somebody else's hope.

The second clock is the pre-holiday pulse, which is a fixed feature of this line and never a surprise. Six to eight weeks before the Chinese New Year, everyone who makes something in China wants it out, and the corridor takes a demand pulse on top of the fourth-quarter one. That is the week when the seam is slow, the yard is full, and the transit time quoted in October is not the transit time you get in January. Put the last pre-holiday shipment down so that it lands with two weeks of slack rather than two days. Two weeks of slack reads like waste in a budget review and reads like wisdom in February.

Now the money, and I will put the assumptions on the table, because otherwise the numbers are decoration. Assume thirty forty-foot containers a month, 80,000 dollars of cargo in each, money at six per cent a year, and this line twelve days faster than the sea option you would otherwise use. Those are placeholders; replace them with your own. Carrying cost is about 13 dollars a container per day, so twelve days is 158 dollars a container. Then assume the rail premium is 1,500 dollars a container. The carrying cost does not come close. If you are selling this lane to a finance director on inventory interest, you will lose that argument, and you should lose it, because the argument is wrong.

What pays for the premium is the thing that never appears on a rate sheet: the promise you made to a customer, and the cost of breaking it. Put a probability on the sea option being late. Assume a 30 per cent chance of a ten-day delay, which is a guess, and it should be your guess rather than mine. That is three expected days, about 39 dollars a container. Add the stockout. If a late delivery costs two per cent of the cargo value in penalties, expediting and a lost season, that is 1,600 dollars on the delayed boxes, and 30 per cent of that is 480. Total expected cost of the sea option, a bit over 500 dollars a container, which is roughly your break-even premium. Above it, rail is a service decision. Below it, rail is a cost decision. Know which one you are making, because somebody will ask.

The larger shipper gets a different answer, and this is where the lane stops being a bet on freight and becomes a bet on a launch date. Assume two hundred containers a month, 150,000 dollars a container, the same six per cent and the same twelve days. Carrying cost is about 25 dollars a container per day, so 300 dollars over twelve days. Two per cent of 150,000 is 3,000 dollars a box, and at a 30 per cent delay probability the expected stockout cost is 900. Add 90 dollars of expected carrying cost and the break-even premium is close to 1,000 dollars a container, which on a 150,000 dollar box is under one per cent of the cargo value. For that shipper the arithmetic needs no hand-waving. It needs a decision about what a launch date is worth.

Here is the point I have not seen in the coverage, and it comes straight out of the load numbers. If the corridor is growing by making trains heavier rather than more frequent, then the constraint you will feel in December is weight, not slots. The axle limit does not care about your commercial urgency. When the yard loads to the limit, the box that is over gets left behind, re-stuffed or re-routed, and that conversation happens at the seam, at night, in a language the shipper is not part of. If you ship dense cargo, your risk this quarter is not that the train is full. It is that your box is the one that gets weighed.

The second missed point is frequency, and it is the one that quietly ruins plans. Thirteen trains a day is the corridor's number. It is not your number. With 76 routes into more than 70 hubs, a great many destinations are served by a branch that runs twice a week, and the branch is where your container actually goes. A shipper who plans on the trunk frequency and gets allocated to a twice-weekly branch has two to four days of variance built into his lead time before anything goes wrong. Ask for the frequency of your destination, in writing, then build the plan on that number rather than on the average.

The third is the seam's own metric, which is dwell rather than speed. What decides your arrival is how long your block sits waiting for a crane slot and a gauge change, and that number moves with the weather, the shift pattern and how many blocks came in that morning. Operators see it. Shippers almost never do. Ask your forwarder for the yard dwell on your last three shipments, in hours, and put it next to what he quoted you. If he cannot produce it, he is quoting the timetable rather than the line.

One direction the conclusion flips is the water. If the sea side settles down in the fourth quarter, the premium on this corridor loses its justification for anything in the middle of the value range, and the shipper who signed a twelve-month commitment at the peak premium spends the year explaining a number he no longer needs. That is not an argument against rail. It is an argument against signing a twelve-month commitment in the quarter when the alternative is broken. Buy the lane by the quarter and keep the option to move volume back. A road you can leave is worth more than a road you are stuck on.

The other direction is a node, and it is more likely than a market move. If the Mongolian section throttles, or the Russian border slows, the corridor does not go from thirteen trains a day to twelve. It goes to nine, because a queue at a seam compounds. Nine a day against a plan built on thirteen is a 21-day transit becoming a 28-day transit, and every promise downstream moves by a week, including the ones you made to people who never asked how the box was getting there.

So I watch three numbers, and none of them is the annual total. Daily departures out of Erenhot, because a drop from thirteen to eleven is the earliest warning you will get. Yard dwell at the transfer, in hours, because that is where a five-day problem hides for a week before it becomes a five-day delay. And the frequency of your own destination branch, because that is the number your lead time is actually built on. Everything else in the announcement is a total, and totals are what you read after the fact.

This road works. I have used it, I have been stuck on it, and I would use it again next quarter. But it is not a motorway. It is a provincial road with two ferries on it, and the ferry timetable decides whether your container arrives on the day you promised. Check the ferries before you leave, carry a spare, and change the tire in the yard rather than at the water. The corridor did 3,500 trips this year because a lot of people did that small unglamorous work. Your box is either one of the reasons it works or one of the reasons it does not.

↑ Back to the key points

— By Ömer Kaya